Mahmudisnan/Tulak-tulak

Description

Tulak-tulak is a traditional writing system associated with the Mandailing people of North Sumatra, Indonesia. Mandailing script is considered a distinct writing system with its own characteristic forms that distinguish it from other traditional scripts in the region. However, its use has become increasingly limited, particularly in written narratives, highlighting the need for documentation and digital preservation. This paper introduces a comprehensive, age-diverse handwritten Mandailing script dataset collected from 197 respondents, primarily representing high school and university students. Each respondent was asked to write each of the 21 character classes at least five times. Following data collection and preprocessing, the dataset comprised 20,685 original images, with 985 images per class, stored in JPG format at a resolution of 512 × 512 pixels. To support model development and evaluation, the dataset was divided into three predefined subsets: training (70%), validation (20%), and testing (10%). The training subset contains 14,490 original and 507,150 augmented images, the validation subset contains 4,137 original and 144,795 augmented images, and the testing subset contains 2,058 original and 72,030 augmented images. The dataset is publicly available on Mendeley Data under a CC BY 4.0 license and is intended for use in OCR model development, Mandailing script digitization. The dataset addresses a significant gap in existing benchmarks, which is expected to support future research and real-world educational applications by facilitating the development of robust and accurate recognition models for Mandailing script.

Repository Structure

Tulak-tulak/
├── Dataset/
│   ├── Original/                         # 20,685 original images (approx. 908 MB)
│   │   ├── train/                        # 14,490 images
│   │   │   ├── Char 01/
│   │   │   │   └── Char_01_<id>.jpg
│   │   │   ├── Char 02/
│   │   │   ├── ...
│   │   │   └── Char 21/
│   │   ├── val/                          # 4,137 images
│   │   │   ├── Char 01/
│   │   │   ├── ...
│   │   │   └── Char 21/
│   │   └── test/                         # 2,058 images
│   │       ├── Char 01/
│   │       ├── ...
│   │       └── Char 21/
│   │
│   ├── Augmentation/                     # 703,290 augmented images (approx. 32 GB)
│   │   ├── train/                        # 492,660 images
│   │   │   ├── Char 01/
│   │   │   │   └── Char_01_<id>__aug_<number>_<technique>.jpg
│   │   │   ├── Char 02/
│   │   │   ├── ...
│   │   │   └── Char 21/
│   │   ├── val/                          # 140,658 images
│   │   │   ├── Char 01/
│   │   │   ├── ...
│   │   │   └── Char 21/
│   │   └── test/                         # 69,972 images
│   │       ├── Char 01/
│   │       ├── ...
│   │       └── Char 21/
│   │
│   └── Dataset.zip                       # Dataset archive
│
└── Uncropped/                            # 1,970 uncropped images (approx. 6.0 GB)
│   ├── 1_1.jpg
│   ├── 1_2.jpg
│   ├── ...                               # Group 1: 985 images
│   ├── 1_985.jpg
│   ├── 2_1.jpg
│   ├── 2_2.jpg
│   ├── ...                               # Group 2: 985 images
│   └── 2_985.jpg
│
└── Tulak_tulak_form.pdf 
└── aug_tulak_tulak_form.ipynb  

Image count summary

Data group	Train	Validation	Test	Total
Original	14,490	4,137	2,058	20,685
Augmentation	492,660	140,658	69,972	703,290
Uncropped	-	-	-	1,970
Grand total				725,945

Data overview

• The classified dataset contains 21 classes, from Char 01 through Char 21.
• The classified data is divided into train, val, and test splits.
• Original images follow the Char_<class>_<id>.jpg filename pattern.
• Augmented images follow the Char_<class>_<id>__aug_<number>_<technique>.jpg filename pattern.
• Each original image has 34 augmented variants.
• The Augmentation directory contains augmented images only; copies of the original images have been removed.
• The Uncropped directory has not been divided into splits or class directories.
• The Tulak_tulak.pdf is the form that is used to collect the dataset
• The aug_tulak_tulak.ipynb provides code to process augmentation for all original images